On the Ways Forward, I think we need to think through what we want from an AI. What's the point of replicating a human? We already have 8 billion of those. What's the point of an AGI? What problem are we trying to solve?
To me, the most obvious need is for domain expertise. We need experts that can assist humans with making decisions. That creates a need for the AI to "explain its work" at any level. In order to be most useful, these systems should be able to reason through scenarios based off a prioritized list of policy goals and engage in What If analysis.
We need the ability to replicate domains of knowledge, experience, and expertise. Think of an automotive AI. We need an AI to be able to learn and replicate that knowledge to other cars. The system collectively learns. Humans should be able to continue to assist with this knowledge collection.
Which leads us into robots in general, since cars are just a kind of robot in this worldview. You want the fleet to be able to get better via their collective experience. You need to be able to identify limitations of processing hardware and storage and adjust what a particular robot may "learn." Ideally, you'd have the ability to make hardware updates to your robot and it would get better.
These are the kinds of things I think we should be working on.
To me, the most obvious need is for domain expertise. We need experts that can assist humans with making decisions. That creates a need for the AI to "explain its work" at any level. In order to be most useful, these systems should be able to reason through scenarios based off a prioritized list of policy goals and engage in What If analysis.
We need the ability to replicate domains of knowledge, experience, and expertise. Think of an automotive AI. We need an AI to be able to learn and replicate that knowledge to other cars. The system collectively learns. Humans should be able to continue to assist with this knowledge collection.
Which leads us into robots in general, since cars are just a kind of robot in this worldview. You want the fleet to be able to get better via their collective experience. You need to be able to identify limitations of processing hardware and storage and adjust what a particular robot may "learn." Ideally, you'd have the ability to make hardware updates to your robot and it would get better.
These are the kinds of things I think we should be working on.